Research to Develop Guidance on Extreme Precipitation Frequency Estimates in Orographic Regions (NUREG/CR-7247)

On this page:

• Publication Information
• Abstract

Download complete document

• NUREG/CR-7247 (PDF – 33.3 MB)

Publication Information

Manuscript Completed: September 2025
Date Published: March 2026

Prepared by:
K. D. Holman
A. P. Verdin 
D. P. Keeney

Bureau of Reclamation
Technical Service Center
Denver Federal Center
PO Box 25007
Denver, CO 80225-0007

Joseph Kanney, NRC Project Manager

Office of Nuclear Regulatory Research
U.S. Nuclear Regulatory Commission
Washington DC 20555-0001

Availability Notice

Abstract

Currently, many engineering design projects use Probable Maximum Precipitation (PMP) to develop Probable Maximum Floods (PMFs), which are then used to evaluate the safety of a proposed or existing facility. Orographic methods employed in hydrometeorological reports (HMRs) used to estimate PMP vary widely through time and by region, if they were employed at all. Deterministic metrics, such as PMP, do not provide decision makers with information on precipitation events less intense than PMP, yet still extreme, nor do they provide information on the expected frequency of such events. The Nuclear Regulatory Commission (NRC) requested assistance from the Department of Interior’s Bureau of Reclamation (Reclamation) in developing guidance on regional extreme precipitation analyses in orographic regions. The methods presented in this report are illustrated across the Tennessee River Valley watershed. The Tennessee River Valley is a region with pronounced orographics, and thus serves as a suitable testbed for the methodologies outlined in this report.

Probabilistic precipitation information often takes the form of a precipitation-frequency relationship, which describes the depth of precipitation and the associated probability of occurrence. NOAA’s Hydrometeorological Design Studies Center produces precipitation frequency estimates across most of the United States. These estimates are published as individual regional volumes of NOAA Atlas 14 (e.g., Bonnin et al. 2006), and include precipitation-frequency estimates as rare as the 1,000-year return period, which corresponds to an annual exceedance probability (AEP) of 1/1,000 or 0.001. However, high-hazard dam safety or nuclear facility applications frequently require precipitation-frequency estimates beyond the 1,000 year return period. This research provides information useful to license applicants on acceptable methods and data sources for estimating and using precipitation-frequency analyses in orographic regions for calculating floods at critical AEPs that the applicants need to consider in evaluating siting factors, conducting Probabilistic Risk Assessments (PRAs), and in designing facilities.

The two main goals of the research were (i) to critically review orographic extreme storm methodologies used in PMP and magnitudes less than PMP in orographic regions; and (ii) to evaluate methodologies for developing precipitation-frequency estimates at AEPs significantly less than those offered in NOAA Atlas 14, in the range of 10^-4 to 10^-6 AEP. To meet the objectives and goals, Reclamation performed the following five major tasks:

1. Review extreme storm precipitation techniques, precipitation-frequency methods, and databases in orographic regions;
2. Develop a methodology to estimate precipitation-frequency analyses in regions of complex topography;
3. Demonstrate the precipitation-frequency methodology and provide uncertainties and confidence intervals at the regional and reactor-site scale for a pilot region in the Tennessee River Valley;
4. Transfer technology to the NRC staff via a training session, including data, software, and scripts; and 
5. Complete a final report that conveys research findings.

The five tasks outlined above were accomplished through the use of existing and new technologies available within the scientific community. Task 1 was accomplished by reviewing the variety of techniques used in orographic storm analyses, including methods from the HMRs, private consultants, and additional methods from the scientific community. Along with the HMRs, we reviewed previous precipitation-frequency data and methods as part of Task 1. We reviewed federal approaches to precipitation-frequency analyses (e.g., NOAA Atlas 14), as well as approaches used by private consultants and the scientific community. We focused primarily on a discussion of regional L-moments and Bayesian inference. 

Tasks 2 and 3 included developing and demonstrating a methodology to estimate precipitation frequency analyses in a region of complex topography. We accomplished these tasks by first combining a known objective clustering algorithm, the Self-Organizing Map (SOM), with two different regional frequency methods, L-moments and  Bayesian inference. These two regional frequency methods vary widely in terms of the level of complexity (and consequently effort) and in the way in which epistemic uncertainty is estimated. The SOM algorithm used a combination of geophysical information (i.e., latitude, longitude, elevation) and observed precipitation data (i.e., mean annual precipitation, mean one-day annual maxima) to identify 14 groups (i.e., homogeneous regions) across the Tennessee River Valley watershed. The two regional frequency methods were applied to historical precipitation observations from these 14 groups located across the Tennessee River Valley watershed. Both analyses developed precipitation frequency estimates using the generalized extreme value (GEV) distribution, a commonly used distribution. Epistemic uncertainty due to probability distribution choice was not the focus of the current research. Results suggest that the SOM algorithm is a useful tool for identifying and grouping similarly-behaved point precipitation observations. Furthermore, the frequency results from these analyses indicate that uncertainty estimates from the L-moments analysis are consistently less than the uncertainty estimates from Bayesian inference. These differences are the result of estimating uncertainty differently between the two methods. L-moments uncertainty estimates are obtained using an ad hoc bootstrap resampling routine; Bayesian inference more properly accounts for uncertainty via Monte Carlo methods, which result in posterior distributions of all model parameters. 

In addition to demonstrating these methods on historical point precipitation observations, we also demonstrated the application of the two regional frequency methods on a gridded precipitation dataset, the 100-member ensemble data from Newman et al. (2015). The application of these two regional frequency methods to a gridded precipitation dataset demonstrates the utility of these methods to an alternative data format. With some modifications, the techniques developed here can be applied to any gridded or point-based precipitation dataset of interest to the users.

Task 4, transfer technology to NRC staff, was accomplished through a meeting between Reclamation personnel and NRC staff that was held at the NRC headquarters in Rockville, MD, on August 1-2, 2017. The event included two days of lectures, with allocated time for questions and discussion. An external hard drive, containing project data, software, and scripts, will be sent to NRC following project closeout. Finally, Task 5 was completed through this report, which conveys all the research findings from the Reclamation team.